Amodei Urges Fast AI Controls to Stop Botnet and Bioweapon Risks

️ Dario Amodei's manifesto on the need to implement terminal control over AI: within 6–12 months artificial intelligence will seize the entire internet with a botnet and paralyze critical infrastructure with damages in the hundreds of billions of dollars — already on the front page of all global media.

Elon Musk from xAI and Sam Altman from OpenAI, despite their complicated relationship with each other, have already agreed with Amodei, who this time is actually extremely specific in the proposed measures: in fact he even published a safety regulation. Since the major US labs, represented by their leaders, have already given consent to regulation, the rapid process of developing standards will now begin above all. Amodei proposes that manufacturers of advanced models agree on it voluntarily among themselves, because government bureaucracy is not flexible, and AI could already start destructive actions, incompatible even with current local wars in terms of damage scale.

An important aspect of the Anthropic CEO's manifesto: he quite clearly identified two main risk vectors in which the destruction of humanity by specific AI methods is indeed possible.

1. The primary near‑term risk from AI is that AI is stronger in intelligence and numbers than old IT‑security staff, so it can easily seize not only a company's private network but also claim control over the internet globally with massive botnet infiltration into companies.

2. Amodei correctly notes that especially strict control must be imposed on AI in the field of dual‑use biological technologies, because AI is fully capable of creating highly contagious and lethal viruses or bacteria that would wipe out all humans.

We must wait to see what response China will give to these initiatives. However, Amodei has very specifically proposed a three‑stage approach to agreements and informal arrangements, where the top priority is strict control over the potential creation of AI‑driven biological weapons, as this is truly not an abstract threat. Chemistry, biochemistry, biology and genetic engineering are fields where AI is potentially even more powerful than in programming, thanks to the convenience of modeling atoms with tokens in transformers.

https://www.washingtonpost.com/business/2026/09/12/anthropic-ai-dario-amodei/5cc3cc44-aec8-11f1-b498-8697f35a6743_story.html
https://edition.cnn.com/2026/09/12/tech/anthropic-ceo-essay-ai